Commit 190cbfa
authored
polygon/sync: message listener to preserve peer events ordering (#10032)
Observed the following issue in a long running Astrid sync on
bor-mainnet:
```
[DBUG] [04-17|14:25:43.504] [p2p.peerEventObserver] received new peer event id=Disconnect peerId=51935aa1eeabdb73b70d36c7d5953a3bfdf5c84e88241c44a7d16d508b281d397bdd8504c934bfb45af146b86eb5899ccea85e590774f9823d056a424080b763
[DBUG] [04-17|14:25:43.504] [p2p.peerEventObserver] received new peer event id=Connect peerId=51935aa1eeabdb73b70d36c7d5953a3bfdf5c84e88241c44a7d16d508b281d397bdd8504c934bfb45af146b86eb5899ccea85e590774f9823d056a424080b763
```
Note the timestamps are the same on the millisecond level, however the
disconnect was processed before the connect which is wrong (connect
should always be first).
This then got the `PeerTracker` in a bad state - it kept on returning
peer
`51935aa1eeabdb73b70d36c7d5953a3bfdf5c84e88241c44a7d16d508b281d397bdd8504c934bfb45af146b86eb5899ccea85e590774f9823d056a424080b763`
as a valid peer to download from, which caused repeated `peer not found`
errors when sending messages to it.
Fix is to have the message listener wait for all observers to finish
processing peer event 1 before proceeding to notifying them about peer
event 2.1 parent 6d9a5fd commit 190cbfa
1 file changed
+12
-1
lines changed| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
170 | 170 | | |
171 | 171 | | |
172 | 172 | | |
173 | | - | |
| 173 | + | |
| 174 | + | |
| 175 | + | |
| 176 | + | |
| 177 | + | |
| 178 | + | |
| 179 | + | |
| 180 | + | |
| 181 | + | |
| 182 | + | |
| 183 | + | |
| 184 | + | |
174 | 185 | | |
175 | 186 | | |
176 | 187 | | |
| |||
0 commit comments